Insurance claims fraud: Assembling an analytics tool box (1 of 3)

Posted on July 20th, 2010 by

In a recent issue of Claims Magazine, I wrote an article called Analytics and claim fraud: Assembling the proper toolbox to prevent and detect scamsIn it, I discuss some strategies that insurers can use to deal with claims fraud. Over the next three weeks, I’ll discuss these strategies in more detail.


Fraud used to be viewed as an unfortunate cost of doing business in the insurance industry, but that attitude is changing. An estimated ten percent of claims turn out to be fraudulent. That, coupled with the development of sophisticated analytical tools, has changed how insurers deal with fraud and fraud investigations.

Types of fraud

On one end of the spectrum, fraud can be simple, isolated and opportunistic, involving low-dollar amounts. On the other end, highly sophisticated fraud rings with links to organized crime can cost insurers significant amounts of money. Generally, very few claim payouts are prevented or recovered as a result of fraud investigations.

Effective anti-fraud programs

The potential to reduce losses due to fraud—not to mention the cost associated with investigating fraudulent claims—is significant. However, establishing an anti-fraud program requires comprehensive thought and planning. Insurers who want to establish an effective anti-fraud program must be mindful that many anti-fraud programs result in false positives from claims that turn out to be non-fraudulent. False positives are a drain on resources, and can have negative effects on customer experience and retention.

As such, effective anti-fraud programs are composed of several sources of information: business rules, anomaly detection, text mining, predictive modeling and social networking analysis. Underlying all these information streams is a process and operating model that is based on analytics.

Good data is paramount

With the emphasis on information, it is critical that insurers have clean, robust and accurate data. A critical look at this data will expose processes and segments that are more prone to fraudulent activity.

This data can take the form of:

  • Background information. This information about a customers includes billing and policy history, contact history and historical claim frequency
  • External data. Sources such as the ISO, CarFax, credit bureaus, medical records and provider information supplement customers’ background information.
  • Previous fraudulent claims. Data indicating past fraud is critical because it provides insight for insurers seeking to develop predictive models to identify future fraud.

Next week, I’ll talk about some of the analytics that can be implemented as part of an anti-fraud program.